198 S. Drucker and M. Puri
estimate private information as a residual and use its correlation with the second stage
dependent variable as a basis for testing whether private information matters.
Clearly, many different approaches can be used to assess the benefits of bank un-
derwriting, and a number of these techniques are utilized in studies of the post-1990
period and in examinations of underwriter fees, which we will discuss later. For ex-
ample, an alternate approach is to use endogenous switching models (see e.g.Fang,
2005; Song, 2004for applications andMaddala (1983)for details on the model). These
models generalize the two-stage approach used inPuri (1996)by allowing commercial
banks and investment banks to have separate yield equations. This relaxes the assump-
tion that the variables that affect yield have the same effect for investment bank and
commercial bank issues. Estimating the model involves two steps. First, the researcher
runs a probit model to determine the probability that the issuer chooses a commercial
bank or investment bank underwriter. In the second step, the researcher estimates two
yield equations separately for investment bank and commercial bank issues, including
independent variables that affect yield as well as the inverse Mills ratio. Interestingly,
when we apply endogenous switching methodology to industrial bonds in the pre-1933
data, similar effects are found.
The evidence that investors paid more for bank underwritten-securities pre-Glass–
Steagall suggests that commercial banks are net certifiers of firm value. However, this
raises an important question of interpretation. Namely, did investors pay more for bank-
underwritten securities because they rationally believed them to be of better quality, or
were investors naïve and banks took advantage of them so that investors paid higher
prices for worse securities? This question can be addressed by examining the ex post
performance of bank underwritten securities.Ang and Richardson (1994), Kroszner and
Rajan (1994), andPuri (1994)examine the ex post performance of securities using data
from the pre-Glass–Steagall period. As noted, if commercial bank-underwritten securi-
ties perform worse than ex ante similar securities that are underwritten by investment
banks, then this would be consistent with commercial banks underwriting securities that
they privately know to be of lower quality. All three studies find no evidence to support
the existence of conflicts of interest.
Ang and Richardson (1994)examine the long-run performance of bonds, using a
comprehensive sample of 647 bond issuances over the years 1926 through 1930. The
authors compare the default rate of commercial bank and investment bank-underwritten
bonds based on the default status of the bonds at two points in time (1934 and 1939)
and find that the default rates are similar for investment bank- and commercial bank-
underwritten securities. While this analysis is limited because the authors do not control
for differences in the characteristics of issuers across the two types of underwriter, the
results suggest that conflicts of interest did not override the certification ability of com-
mercial banks.
Kroszner and Rajan (1994), using data from the first quarters of the years 1921
through 1929, examine the relative performance of industrial bonds that are underwrit-
ten by commercial banks with those that are investment bank-underwritten. The main
measure of bond performance is the default rate because reliable price data is scarce